Vous êtes sur la page 1sur 12

-T

c H A P TE R

74

CASE SYN C R E TIS M


MATTHEW B A ERM AN

r 4 . L I N r R o D u cr r o N
syncretisrrwe understand the combination of multiple distinct case By the term case in a single form. Distinct casevalues are determined on a language-specihc values basis, so that casesyncretism by this definition involves an observable asymmetry paradigms within a language.'In the most obvious pattern, multiple case between forms in one paradigm correspond to a single caseform in another. For example, in the singular paradigrn of a-stem nouns has eight distinct caseforms, which Sanskrit, justifies In the non-singular paradigns, however, many positing eight distinct cases. of thesedistinctions are collapsed,particularly in the dual. (r) Sanskrit a-stem noun (\\4ritney 1889:tt4-r5) god'
NoM voc ACC rNS DAr ABL cEN Loc

singular
devas deva devam devena devaya devat devasya deve

dual

plural

devan devais --- l rdeveDnyas i devebhyas l devanam devesu

I One might also identi$' an asymmetry along the diachronic dimension, where n'rultiple ctrses in someolder stageofa language correspond to a single casein a more recent stage (e.g. the various

Y
M A T T H EW BAERM AN

Alternatively, distinct casesmay be identifiable solely on distributional grounds (Zaliznjak ry73; Comrie r99r). For example, in ClassicalArmenian, the accusative never has a distinct form, being syncretic either with the nominative (singular nouns) or the locative (plural nouns, and pronouns). (z) ClassicalArmenian (Meillet ry36:8t-2,9t) 'father'
NOM A CC

(z

'fathers'

'I'
CS f.--__-l

Loc cEN DAT ABL rNS

hawr hawr hawr hawr harb

I hars I
harc' harc' harc' harbk '

E;] tl

hark'

lrs I tl
lr r
l:^

im inj inen inew

St

Some apparent examples of casesyncretism are doubtlessly superficial, and are best treated as accidental homophony. Thus in Latvian (Indo-European, Baltic), some declension classeshave a nominative singular ending -s, and some have a genitive singular -s. The two setsoverlap, so that there are declension classes which have both nominative and genitive singular in -s, resulting in syncretism of these two cases. Howeveg since the s -endings occur independently of each other, there is no compelling reason to posit a systematic relationship between the two identical forms. (:) Latvian (Forssman2oo1:113, tr4, rr8) a. i-stem 'brother' NoM sc cEN sc bralis brala b. a-stem c. u-stem 'market' 'hand' roka rokas

Id

CC

n t (-

AI

sl-

Fr

Ir

th

\)

However, there is also considerableevidencethat casesyncretism maybe systematic. For example, in Sanskrit nouns, the ablative and genitive singular are distinct only for a-stems (seedevasabove (r)). For all other noun classes they are syncretic, and the actual form may vary, as illustrated in (4).

VJ

St m

AS oblique casesin (r) correspond to a single oblique casein many modern Indic languages).In this chapter, we restrict our focus to synchronically identifiable phenomena.

sy

stl

-7C A S E S Y N C R E TIS M 221

(+) Sanskrit (\A4ritney1889: u9, r39)


i- s t e m ' fi re ' NoM voc ACC rNS DAr agnis agne agnim agnina agnaye r-s l e m ' g i v e r' dAta datar datara datra datre

ABL
cEN
Loc

F*q
ia g nesl
agna u

ldaturl
d a ta ri

Here it is clear that the syncretism is not accidental, as it is repeatedacross multiple exponents:the i-stems have the ending -es, while the r-stems have -ubefore the stem-final consonant.

r 4 .2

P OS S TB LE

TNTE RP RE T A T T O NS

Identifzing a syncretic pattern as systematic still leavesopen the question of which component of grammar encodes it, morphosyntax or morphology. on one interpretation, syncretic caseforms reflect some unity at the level of feature structure, an approach which owes a particular debt to Jakobson (tg|e). Thus, we might supposethat individual case values were in fact bundles of more abstract values. For example,we might say that 'ablative' is really a cover term for the feature bundle [+x, -y], while'genitive'is [+x, +y] . Syncretismoccurs in paradigms where only th es har ed v alue I rx l i s a c l i v e . (i) Possiblefeatural decomposition a-stem i-stem, r-stem

anr [+x, +y] crN [+x, -yJ

[+x, +y] [+x, -yi

[+x] [+x]

Viewedin this way' casesF:rcretismis a window into the otherwise covert internal structure of case features. A second approach is to treat syncretism as purely a morphological phenomenon: the form of the ablative and genitive are stipulated as being identical, without reference to their function. on this interpretation, syncretism demonstratesthe independence of morphosyntactic and morphological structure,and hence the autonomy of morphology.

222

M A T T HEW

BAERM AN

For any individual language, it is hard, if not impossible, to fully justify one or the other approach on a principled basis. However, from a broader crosslinguistic perspective,each approach makes certain predictions. If casesyncretism is motivated by shared meaning or function, then we should be able to predict which cases should be syncretic with each other, and find these repeated across languages.If, on the other hand, case syncretism is purely morphological, then we would not be able to make any cross-linguistic generalizations in terms of case function. As one might expect, the truth appears to lie somewhere in between.

L4.3 Tv p o L o G Y
The tendencies in case syncretism are clearest if we divide case syncretism into four types, numbered in decreasingorder of their cross-linguistic frequencies.The figures are based on Baerman, Brown, and Corbett (zoo5), which is a survey of a controlled sample of ry9 languages. Of these, seventy-four display inflectional marking of case on nominals. Forty of these languages have some sort of case syncretism: thus case syncretism is found in roughly half of the languages with inflectional casemarking. . Type r: syncretism of the core cases, i.e. the cases expressingthe grammatical roles of subject (transitive and intransitive) and object. This is found in thirty out of the forty languageswith casesl.ncretism (of the remaining ten, two do not distinguish inflectionally at all). the core cases . Type z: syncretism of a core casewith some non-core case.Typically it involves what some might call the 'marked' core case(accusativeor ergative,as opposed to nominative or absolutive). This is found in twenty-seven of the forty languages, often alongside type r. . Typ" 3: s1'ncretismbetween non-core cases. This is found in five languagesout of the forty, always alongside type 7 or 2. . Type 4: casesyncretism compounded with some other feature, such as number, found in five languagesin the sample. These languagesall display type 1 or 2 as well. While the ligures must be treated with some caution (since the identilication of case syncretism is open to individual interpretation), they do suggestthat syncretism involving core casespredominates. Indeed, the presence of case syncretism in a language implies the presence of syncretism involving the core casessomewhere in the system.

C A S E S Y N C R E TIS M

221

r4.j.r TYp. t
An example of nominative/accusative syncretism is given above in (z). Example (O) illustrates absolutive/ergative syncretism in Basque, which affects the proximate plural of all nouns (as well as the plural of some pronouns, e.g. zuek 'you.rr.ens/enc'). (6) Definite plural in Basque (Hualde zoo3: ry3)

plural general 'the places'


ABS lekuak ERG lekuek DAr lekuei

proximateplural 'the places here'

The cross-linguistic frequency of type r is perhaps not surprising when we consider it within the broader context of semantic and syntactic conditions that affect the marking of core arguments. For example, the notion of differential object marking describesthe tendency for languagesto restrict overt object marking (accusative case)to certain arguments, while conversely, the notion of differential subject marking describeshow languagesrestrict overt transitive subject marking (ergative Both tendencies can be seen in Dhangu (Pama-Nyrngan, Yolngu) where the case). is marked on personal pronouns and the ergativeis marked on inanimate accusative nouns; animate nouns (in particular, human) fall in-between, marking both cases. We can group the two tendencies together under the rubric'differential argument marking'. (z) Core casesyncretismin Dhangu (Schebeck r976:5t7,526-7) personal pronoun
(t()

animate noun
Woman

inanimate noun
SIOry

We.DU.EXCL

E RG-INS ABS

taykkayu taykka
gahnrunYa

ACC

taykkanra

The idea behind differential argument marking is that there are prototypical patientsand prototypical agents,the criteria being semantic or pragmatic, prototypical objectsbeing inanimate, indefinite, and non-topical, prototypical agentsbeing animate,definite, and topical. However, it is often the case that purely morphologicalor lexical categoriesplay a role as well. For example, in Dhangu, although animacy plays the key role, there is a sharp distinction between pronouns and nouns, regardlessof animacy. In Luiseflo (Uto-Aztecan, Takic), inanimate nouns

224

M AT T HEW

BAERM AN

ending that animatenouns have,in keeping with differential lack the accusative of regardless doesmark the accusative, But an attributiveadjective objectmarking. the animacyof the noun it is modif ing. (8) Luiseflo (Mal6cot t963:zo8)
inanimate animate = 'bad man' grass+ bad = 'bad grass' man + bad NoM ACC ja?6i ja?dtSr alaxwr5 alaxvrrrtSt alaxwrd alaxrantSr

More extreme examples of the morphologization of such patterns are familiar from Indo-European languages.Thus, in Latin, nominative and accusativeare not distinguished by neuter nouns. While this would seem to follow from the notion of differential object marking (neuters are inanimates, and thus prototypical patients), the relationship between animacy and gender is not direct: inanimates are also well representedamong the masculine and feminine genders,which do distinguish the two cases.

14.3.2 'Iyp"

"

In a nominative-accusative system, type 2 tlpically involves syncretism of the accusative with some non-core case, while in an ergative-absolutive system, it is the ergative which is involved. In Martuthunira (Pama-Nnrngan), the accusative and genitive are sy'ncretic for all nominals except vowel-stem nouns. (S) Martuthunira (Dench ry95: 65, to7)

V-stem 'money,rock'
NOM ACC GEN LOC

N-stem 'tree' kalyaran

L-lrr-stem pronoun 'daughter' 'who' kurntal


ngana

kanyara kanyaraa kanyarawu kanyarala

*t"b",l try"'""E F" Fs"-C;l


I lnganangu I lkalyaranku I lkurntalyu
kalyaranta kurntala nganala

for all and genitiveare syncretic spokenin Bolivia), ergative In Araona (Tacanan, like the plural pronouns). singularpronouns (nounsbehave nominalsexcept (Pitman (ro) Pronouns r98o: 8z) in Araona
Is c
ABS ema

l ru
tseda

lpr cuama

ERG yama f;Al cE N quima ltr"udu I lcuam a (-ja )l

C A S E S Y N C R E TIS M

22\

often the syncreticform appearsto be an extensionof the range of the noncorecaseform. For example,in chukchi (chukotko-Kamchatkan), the ergative (-instrumental) and locativecases are syncreticin the plural, and the form itself appears to resultfrom an extension of the locativeending-k to the ergative. (The suffix-r(a)- marksthe obliqueplural; the schwa - skorik t96r: zo4.) is epenthetic (tr) Chukchithird declension noun (Skorik ry6r: zq-4) 'child'
ABS ERG-INS Loc ABL DAT oRr COM

singular
nenena nenene-te nenene-k nanana-jpa nanana-ZtA nenene-7jit Za-nanana-ma

plural
nenene-t

nanana-r-7apc nanana-ra-ka nenene-ra-/it /a-nanana-ro-ma

This may be combined with other patterns of syncretism yielding a situation where the marked core casehas no dedicated form, and is only identifiable on distributional grounds. For example, in (rz) and (r3), type 2 s)'ncretismis in complementary distribution with type r syncretism, in accusativeand ergative systems,respectively. (rz) Estonian

'book'
NOM
ACC GEN

singular
raamat

plural

raamatute
raamatut

PART

raamatuid

(U) Lak (Zirkov ry55:36,64-6)


noun'house'
ABS
ERG GEN

pronoun'I'

k'atta

DAT k'atlu-n

ttu-n

such examplesmight be interpreted as a kind of differential argument marking, with the proviso that there is no dedicated form for the marked core case.Not all examples are of this type, though: for example, in Bonan (Mongolic), the accusative is always distinct from the nominative; however, the accusative itself is always syncretic with some non-core cases,depending on word class (with the genitive for nouns, and with the dative-locative for pronouns).

r I I
226 M A TT HEW BAERM AN

-=-

(r+) Bonan (Todaeva r99T 35)


noun'foliage'
NOM GEN ACC D A T - L OC ABL I N S _COM

pronoun'he' ndLaq ndZag-ne

labtog

I Ilabiou-ne
labio4-de Iabioq-se labior3-yale
ndZag-se ndi.aq-yale

Of course, for Bonan, an alternative analysis is possible: one might say that they reflect a quirky system of case assignment rules, whereby nominal direct objects were assigned genitive case and pronominal direct objects were assigned dativelocative. As always, there is a trade-off: the simpler the morphological rules, the more complicated the syntactic rules. In some cases,however, it is clear that the type z pattern is morphological and not syntactic. For example, in Russian, some nominals display an alternation of type r and type z syncretism according to animacy: the accusativeof inanimates is identical to the nominative, while the accusativeof animates is identical to the genitive. The pattern is mirrored on agreeing adjectives. (rl) Inanimate - animate alternation in Russian inanimate'oldchair'
NOM ACC GEN

animate'oldperson' staryj ielovek starogo ielove staroso celoveka

f-, 't"tl staryjstol


I

] starogostola

However, this alternation is partly determined by inflectional class.This is evident when we contrast the behaviour of a-stem nouns with that of agreeing adjectives. The a-stems are an inflectional classwhich has the non-syncretic accusativecase ending -u in the singular. The use of this ending is not sensitive to animacy. But if the noun is an animate masculine, the adjective will still display the syncretic pattern illustrated in (rS): (16) Animate a-stem with adiective 'old man' NoM
ACC GEN

staryj

muZiina muZiinu muzclny

Thus, the pattern is morphological, not syrtactic: although the adjective takes the form of the genitive, it is clearly functioning as an accusative,as evidenced by the

C A S E S Y N C R E TIS M

227

marking on the noun. This is similar to what is found in the Luiseio paradigm case in (8), where only the noun, but not the adjective, displays animacy-conditioned syncretism.

14.3.3 Typ" r
Giventhe greater heterogeneity of the non-core case distinctions across different languages, one should not expect to be able to make precisecomparisons. Nevertheless, some general patterns can be found. One important observation concerns the sort of exuberant patterns found in Indo-European, such as those in Sanskrit (see (r) above).Though one might be tempted to seethis as a reflection of an underlying networkof morphosyntactic relationships, the fact is that these patterns are practicallylimited to Indo-European - unlike type r and type 2 syncretism, which are widelyfound in other languagefamilies. Isolated examplesdo occur, for example in ErzjaMordvin the dative and illative are syncretic in the singular of definite nouns. (That thesetwo functions should be combined is perhaps unsurprising; however, thatit shorild be limited to the definite singular forms is curious.) One pattern which Indo-European does share with non-Indo-European languages is the collapse of all non-core cases,or of all non-nominative (nonabsolutive) into a single'oblique' form. Compare the paradigm of feminine nouns in Sogdian(Middle Iranian) in (r7), with the Georgian attributive demonstrative i n ( r 8a) . (g) Sogdian(Sims-Williams r98z: 67) masculine 'person'
NOM ACC GEN LOC ABL raml TAlnU TAInC ramya TANA

feminine 'tree'

(r8) Georgian (Hewitt 1996:97-8) a. 'this tree'


NOM ERG DAT GE N INS A DV

b. 'this one' aman


AMAS

11
lu- I

xem
XCS

xis lu-| xit lu- xed lam l

amis amit amad

->-

228

MAT T HEW

BAERM AN

Note that, in spite of the superficial similarity the nature of the patterns is quite different. In Sogdian, the syncretism is a consequence of the syncretic ending -ya. In Georgian, the s1'ncretismis a consequenceof thelack of an ending. The paradigm of attributive demonstratives in prenominal position is a reduced version of the full paradigm. In other contexts, e.g. as a free-standing pronoun (r8b), the demonstrative shows full caseinflection, alongside a stem alternation between the nominative In prenominal position, it retains the stem alternation but loses and the other cases. the caseendings.

14.3.4Type +
simpleexamplecomesfrom A relatively of differentpatterns. \p" + is a grab-bag two caseforms (direct versusoblique) and Pashto(Iranian), which distinguishes the obliquesingularis two numbers(singularand plural). For somenoun classes, identicalto the direct plural. (rS) Pashto(Grjunbergand Edel'manry87:49-50) 'person' singular plural direct oblique
There have been some attempts to find a systematic principle in these patterns basedon markednessrelations between the cumulated morphosyntactic values (e.g. Serzisko r98z). For example, we might say that sari always realizes markedness disharmony between the case and number values: marked case (oblique) in the unmarked number (singular) or unmarked case (direct) in the marked number (plural). However, there is no consistent evidenceto support this interpretation. Indeed, there are examples such as the comitative singular/locative plural syncretism of Saami (Hansson 2ooz; see below) which hardly lend themselves to a coherent analysis under anyone's conception of markedness,which suggeststhat this is the wrong line of inquiry.

'farmer' singular plural

r4.4 D r e . c HRO NY
it may well be Ratherthan seeksomekind of unified explanationof syncretism, processes. The kinds of diachronic product various it the of more fruitful to see as The most widespread data suggest that we needto considerat leastfive processes.

C A S E S Y N C R E TIS M

229

type of casesyncretism, that of the core cases,may in many instances represent the outcome of desyntacticization, that is, the morphologizedrelic of what was once an active syntactic rule. Thus, differential argument marking, in its purest form, representsa slmtactic rule conditioned by semantic factors. The more morphological factors (such as inflection class) come into play, the less tenable a purely slntactic analysisbecomes' to the point where one has a system where casechoice is consistentacross semantic classes, and it is morphology which determines patterns of formal identity. Of course, where to draw the line between a slmtactic or morphological analysis of a given example may be a matter of some controversy. Accidental sFrcretism occurs as a by-product of phonological or morphological changes. For example, the collapse of caseforms due to phonological reduction or loss is a familiar theme in the history of Indo-European languages,as in Middle High German, where the reduction of unstressed vowels in final syllables to schwa (graphically e) in Middle High German led to widespread mergers throughout the inflectional system: compare old High German bofo Nou.sc - botun ACC.sc - botin cpN/oer.sc 'messenger' to Middle High German bote NoM.sc - boten 'ecc/cpN/rer.sc' (Paul, wiehl, and Grosse r9g9: zor). The accidental effects of a morphological change are apparent in Russian, where the feminine instrumental singular ending -ojuhas become -oj over the last fewcenturies (which was not a regularphonological development). In adjectivesand pronouns, where the ending -of is also found for other cases, this led to syncretism, while in nouns it did not. (zo) Change in the instrumental singular in Russian adjective'old' NoM ACC cEN DAr Loc INS staraja staruju staroj staroj staroj staroju --> staroj noun'book'

kniga knigu knigi knige knige knigoju -+ knigoj

Thatis, a morphologicalchange affecting the shape of the endingled to syncretism in adjectives and pronouns, but the syncretismitself wasnot the motivating factor. However, accidentalpatterns of syncretismmay well be interpreted by language users assystematic, as evidencedby their analogicalextension.For example,polysyllabic nominals in Saamiregularly display syncretismof the comitative singular and locativeplural, a pattern which clearly has no functional motivation, and seems to havebeenthe accidental result of purely phonologicaldevelopments. In mostdialects, theseforms are distinct for pronouns.However,in somenorthern dialects, this purely formal pattern of syncretism hasbeenextended by analogyto monosyllabic nominals,through extension of the comitativesingularform to the locative plural (Hansson zooT).

-rrF
M AT T HEW BAERM AN

Alternatively, syncretism may arise from changesin the treatment of casefunctions. For example, in the early stagesof Middle Indic (which descendedfrom a system like that of Sanskrit), the genitive form had replaced the dative form in all paradigms except the singular of a-stem nouns (Pischel r98r: 289). This syncretism was not a result of phonological change; rather, there was a reassignment of case functions. The tasks originally allotted to the dative were reassignedto the genitive in most paradigms. Such reanalysesmay ultimately lead to a shrinking or disapfor example, in many modern Indo-European pearanceof casedistinctions, as Seen, (seefn. r). languages While the preceding three types involve the merger of originally distinct forms, syncretism can also conceivably result from the partial bifurcation of what was originally a single form. This is the scenario that Dench (1998) reconstructs for *-ku oiginally served for a single Martuthunira (see (9) above). The ending accusative-genitivecase.In its role as a genitive, it could be followed by other case endings (double caseor casestacking), which created a distinct set of phonological * and morphotactic conditions which allowed -ku to develop separatelyin the two case environments, at least for vowel-stem nouns. As a result' we can now speak of two distinct cases in Martuthunira, accusative and genitive' Where they are syncretic, this continues the original, undifferentiated state of affairs.

1 4 .5 GONCLUSTON
In principle, casesyncretism may represent either of two things: (i) an underlying semantic or functional unity, such that the syncretic caseform can be said to realize what is, at some level, a single case, or (ii) a morphologically defined identity of form. Clear examples of (i) may be taken to demonstrate interrelationships between the semantic or functional components of case,while clear examples of (ii) may be taken to demonstrate the autonomy of morphology. In practice, it is frequently diffrcult to make conclusive argument for either interpretation. Given that diachronic shifts between the two undoubtedly occur (as sketched in Sr+'+)' this is not surprising. What this means is that not only is it important to recognize the dual significance of casesyncretism, it is also important to remain circumspect in one's interpretation of it.

Vous aimerez peut-être aussi